[engine.predictor] Allow prediction to work when batch-size!=0; Discard CSVs and use JSON for output; Implement lightning logging like in training; Separates batched (collated) data at prediction; Adapts all models to new paradigm; Remove CSVPredictionsWriter; Adapt predict script to all changes

[ptbench.engine.device] Move AcceleratorProcessor to engine submodule, revamp it, and reflect changes in trainer

[ptbench.engine] Simplified, documented and created type hints for the ``callbacks`` module; Added type hints to the resource module

Daniel CARRON authored 1 year ago and André Anjos committed 1 year ago

Implemented BCEWithLogitsLoss reweighting function
Removed save_hyperparameters from model
Apply augmentation transforms on singular images
Fixed model summary

The CUDA_VISIBLE_DEVICES environement variable is set if a device is
specified or we raise an error if it has not been set.
Code for accelerator handling has been moved to its own utils script.